Scaling modes

A cluster runs in one of these scaling modes. For guidance on choosing and configuring them, see Cluster sizing and scaling.

Mode	How it behaves	Best for
Fixed	Runs exactly the chosen executor count at all times (XS–XXL or custom 1–100)	Steady, predictable load
Autoscaling	Scales between a min and max executor count based on load	Variable or spiky load
QPS-based	Auto-sizes planner replicas to a target queries-per-second	A known target QPS
Scale-to-zero	`min = 0`; drops to zero replicas when idle (with auto-suspend)	Long idle periods, batch, dev/test

See also